yetrun's picture
ver3: 将源码迁入 src/deep_learning 包,重塑训练流水线,规范 data/model 契约
07cb7d3
Raw
History Blame Contribute Delete
724 Bytes
"""Wiki 数据集 Runner
Usage:
python data/wiki/runner.py test_dataset
ENV=production python data/wiki/runner.py test_dataset
"""
import pathlib
import sys
sys.path.insert(0, str(pathlib.Path(__file__).parent.parent.parent))
from deep_learning.data.runner import DatasetRunner
from .dataset import WikiDataset
from deep_learning.env.resolve import resolve_path, resolve_env
dataset = WikiDataset(
data_dir=str(
resolve_env(resolve_path("data/dev/mini_c4"), resolve_path("~/data/wiki/mini_c4"))
),
tokenizer_type=resolve_env("character", "sentence_piece"),
sequence_length=256,
)
runner = DatasetRunner(
dataset=dataset,
name="wiki",
)
if __name__ == "__main__":
runner()